opencode/packages/core/test/plugin/models-dev.test.ts

1198 lines
47 KiB
TypeScript

import path from "path"
import { describe, expect } from "bun:test"
import { Money } from "@opencode-ai/schema/money"
import { Context, Effect, Exit, Layer, Scope } from "effect"
import { TestClock } from "effect/testing"
import { Catalog } from "@opencode-ai/core/catalog"
import { Integration } from "@opencode-ai/core/integration"
import { AppNodeBuilder } from "@opencode-ai/core/effect/app-node-builder"
import { LayerNode } from "@opencode-ai/util/effect/layer-node"
import { Bus } from "@opencode-ai/core/bus"
import { Location } from "@opencode-ai/core/location"
import { Model } from "@opencode-ai/core/model"
import { ModelsDev } from "@opencode-ai/core/models-dev"
import { ModelsDevPlugin } from "@opencode-ai/core/plugin/models-dev"
import { Plugin } from "@opencode-ai/core/plugin"
import { PluginHost } from "@opencode-ai/core/plugin/host"
import { ProviderPlugins } from "@opencode-ai/core/plugin/provider"
import { Provider } from "@opencode-ai/core/provider"
import { AbsolutePath } from "@opencode-ai/core/schema"
import { withEnv } from "../fixture/env"
import { location } from "../fixture/location"
import { testEffect } from "../lib/effect"
import { catalogHost, host, integrationHost } from "./host"
import { PluginTestLayer } from "./fixture"
const locationLayer = Layer.succeed(
Location.Service,
Location.Service.of(location({ directory: AbsolutePath.make(import.meta.dir) })),
)
const layer = AppNodeBuilder.build(LayerNode.group([Catalog.node, Integration.node, Bus.node]), [
Location.node.replace(locationLayer),
])
const it = testEffect(layer)
const real = testEffect(PluginTestLayer)
const isolated = testEffect(Layer.empty)
const models = (file: string) =>
AppNodeBuilder.build(ModelsDev.node, [ModelsDev.node.replace(ModelsDev.configured({ file, fetch: false }))])
function required<T>(value: T | undefined): T {
if (value === undefined) throw new Error("Expected value")
return value
}
// One complete Location graph behind the production plugin host. Two of these stand in for
// two Locations that share a single models.dev snapshot instance.
const owner = Effect.gen(function* () {
const context = yield* Layer.build(PluginTestLayer)
const host = yield* PluginHost.make(Context.get(context, Plugin.Service)).pipe(Effect.provideContext(context))
return {
context,
host,
bus: Context.get(context, Bus.Service),
catalog: Context.get(context, Catalog.Service),
integration: Context.get(context, Integration.Service),
}
})
// Every nested overlay shape the normalized snapshot can carry, so in-place mutation of any
// existing nested value is observable on the source object.
const richSnapshot = (name = "Acme") => {
const providerID = Provider.ID.make("acme")
const modelID = Model.ID.make("gpt-5.4")
const snapshot = [
{
info: {
id: providerID,
name,
activation: "auto",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
settings: { baseURL: "https://api.acme.test/v1", thinking: { type: "adaptive", display: "summarized" } },
headers: { "x-acme": "provider" },
body: { service_tier: "default", tags: ["stable"] },
},
environment: ["ACME_API_KEY", "ACME_HOST"],
models: [
{
id: modelID,
modelID,
providerID,
name: "GPT-5.4",
family: Model.Family.make("gpt"),
settings: { baseURL: "https://models.acme.test/v1", reasoning: { effort: "low" } },
headers: { "x-mode": "fast" },
body: { service_tier: "priority", options: { top_k: 1 }, stop: ["<end>"] },
capabilities: { tools: true, input: ["text"], output: ["text"] },
variants: [
{
id: Model.VariantID.make("low"),
settings: { thinking: { type: "adaptive", display: "summarized" }, effort: "low" },
body: { max_tokens: 1024 },
},
],
time: { released: Date.parse("2026-01-01") },
cost: [
{
input: Money.USDPerMillionTokens.make(2.5),
output: Money.USDPerMillionTokens.make(15),
cache: { read: Money.USDPerMillionTokens.make(0.25), write: Money.USDPerMillionTokens.zero },
},
{
tier: { type: "context", size: 200_000 },
input: Money.USDPerMillionTokens.make(5),
output: Money.USDPerMillionTokens.make(22.5),
cache: { read: Money.USDPerMillionTokens.make(0.5), write: Money.USDPerMillionTokens.zero },
},
],
status: "active",
enabled: true,
limit: { context: 1_050_000, input: 922_000, output: 128_000 },
},
],
},
] satisfies readonly ModelsDev.Snapshot[]
return { providerID, modelID, snapshot }
}
describe("ModelsDevPlugin", () => {
isolated.effect("shares one snapshot between Locations while each catalog mutates only its own copies", () =>
Effect.gen(function* () {
const { providerID, modelID, snapshot } = richSnapshot()
const pristine = JSON.stringify(snapshot)
const source = ModelsDev.Service.of({ get: () => Effect.succeed(snapshot), refresh: () => Effect.void })
const first = yield* owner
const second = yield* owner
for (const each of [first, second])
yield* ModelsDevPlugin.effect(each.host).pipe(
Effect.provideService(ModelsDev.Service, source),
Effect.provideContext(each.context),
)
// The stored environment method must own its names array; the source array is shared.
let names: readonly string[] | undefined
yield* first.integration.transform((draft) => {
names = draft.method
.list(Integration.ID.make(providerID))
.flatMap((method) => (method.type === "env" ? [method.names] : []))[0]
})
expect(names).toEqual(snapshot[0].environment)
expect(names).not.toBe(snapshot[0].environment)
// Catalog-owned provider records are copies, not the source's nested objects.
const stored = required(yield* first.catalog.provider.get(providerID))
expect(stored.settings).toEqual(snapshot[0].info.settings)
expect(stored.settings).not.toBe(snapshot[0].info.settings)
expect(stored.headers).not.toBe(snapshot[0].info.headers)
expect(stored.body).not.toBe(snapshot[0].info.body)
// A later plugin in the first Location mutates existing nested values in place through the
// production host, the way the Bedrock, Azure, and config provider plugins do.
const scope = yield* Scope.make()
yield* first.host.catalog
.transform((catalog) => {
catalog.provider.update(providerID, (provider) => {
required(provider.settings).baseURL = "https://override.acme.test/v1"
required(provider.settings).thinking.type = "disabled"
required(provider.headers)["x-acme"] = "override"
required(provider.body).service_tier = "priority"
required(provider.body).tags.push("override")
})
catalog.model.update(providerID, modelID, (model) => {
required(model.settings).reasoning.effort = "high"
required(model.headers)["x-mode"] = "slow"
required(model.body).options.top_k = 7
required(model.body).stop.push("<stop>")
const variant = required(model.variants[0])
required(variant.settings).thinking.type = "disabled"
required(variant.body).max_tokens = 4096
model.capabilities.input.push("image")
required(model.cost[0]).cache.read = Money.USDPerMillionTokens.make(9)
required(required(model.cost[1]).tier).size = 1
required(model.cost[1]).input = Money.USDPerMillionTokens.make(42)
model.limit.context = 1
model.time.released = 5
})
})
.pipe(Scope.provide(scope))
const mutatedProvider = required(yield* first.catalog.provider.get(providerID))
const mutated = required(yield* first.catalog.model.get(providerID, modelID))
expect(mutatedProvider.settings).toEqual({
baseURL: "https://override.acme.test/v1",
thinking: { type: "disabled", display: "summarized" },
})
expect(mutatedProvider.headers).toEqual({ "x-acme": "override" })
expect(mutatedProvider.body).toEqual({ service_tier: "priority", tags: ["stable", "override"] })
expect(mutated.settings).toEqual({
baseURL: "https://models.acme.test/v1",
thinking: { type: "disabled", display: "summarized" },
reasoning: { effort: "high" },
})
expect(mutated.headers).toEqual({ "x-acme": "override", "x-mode": "slow" })
expect(mutated.body).toEqual({
service_tier: "priority",
tags: ["stable", "override"],
options: { top_k: 7 },
stop: ["<end>", "<stop>"],
})
expect(mutated.variants).toEqual([
{
id: Model.VariantID.make("low"),
settings: { thinking: { type: "disabled", display: "summarized" }, effort: "low" },
body: { max_tokens: 4096 },
},
])
expect(mutated.capabilities.input).toEqual(["text", "image"])
expect(mutated.cost[0]?.cache.read).toBe(Money.USDPerMillionTokens.make(9))
expect(mutated.cost[1]).toMatchObject({ tier: { type: "context", size: 1 }, input: 42 })
expect(mutated.limit.context).toBe(1)
expect(mutated.time.released).toBe(5)
// The sibling Location and the shared source are untouched.
const sibling = required(yield* second.catalog.model.get(providerID, modelID))
expect(sibling.settings).toEqual({
baseURL: "https://models.acme.test/v1",
thinking: { type: "adaptive", display: "summarized" },
reasoning: { effort: "low" },
})
expect(sibling.headers).toEqual({ "x-acme": "provider", "x-mode": "fast" })
expect(sibling.body).toEqual({
service_tier: "priority",
tags: ["stable"],
options: { top_k: 1 },
stop: ["<end>"],
})
expect(sibling.variants).toEqual(snapshot[0].models[0].variants)
expect(sibling.capabilities.input).toEqual(["text"])
expect(sibling.cost).toEqual(snapshot[0].models[0].cost)
expect(sibling.limit).toEqual(snapshot[0].models[0].limit)
expect(sibling.time).toEqual(snapshot[0].models[0].time)
expect(required(yield* second.catalog.provider.get(providerID)).settings).toEqual(snapshot[0].info.settings)
expect(JSON.stringify(snapshot)).toBe(pristine)
// Removing the mutating transform rebuilds the first catalog from the shared source.
yield* Scope.close(scope, Exit.void)
expect(yield* first.catalog.model.get(providerID, modelID)).toEqual(sibling)
expect(required(yield* first.catalog.provider.get(providerID)).settings).toEqual(snapshot[0].info.settings)
expect(JSON.stringify(snapshot)).toBe(pristine)
}),
)
isolated.effect("replaces its snapshot reference on refresh without touching the sibling Location", () =>
Effect.gen(function* () {
const initial = richSnapshot("Acme")
const refreshed = richSnapshot("Acme Refreshed")
const pristine = JSON.stringify(initial.snapshot)
const current = { snapshot: initial.snapshot }
const source = ModelsDev.Service.of({
get: () => Effect.sync(() => current.snapshot),
refresh: () => Effect.void,
})
const first = yield* owner
const second = yield* owner
for (const each of [first, second])
yield* ModelsDevPlugin.effect(each.host).pipe(
Effect.provideService(ModelsDev.Service, source),
Effect.provideContext(each.context),
)
expect(required(yield* first.catalog.provider.get(initial.providerID)).name).toBe("Acme")
current.snapshot = refreshed.snapshot
yield* first.bus.publish(ModelsDev.Event.Refreshed, {})
// Integration and catalog reloads are debounced sequentially.
yield* TestClock.adjust("500 millis")
yield* TestClock.adjust("500 millis")
yield* TestClock.adjust("500 millis")
expect(required(yield* first.catalog.provider.get(initial.providerID)).name).toBe("Acme Refreshed")
expect(required(yield* second.catalog.provider.get(initial.providerID)).name).toBe("Acme")
expect(JSON.stringify(initial.snapshot)).toBe(pristine)
expect(JSON.stringify(refreshed.snapshot)).toBe(JSON.stringify(richSnapshot("Acme Refreshed").snapshot))
}),
)
real.effect("keeps the retained model seed unchanged across catalog replay", () =>
Effect.gen(function* () {
const catalog = yield* Catalog.Service
const plugins = yield* Plugin.Service
const providerID = Provider.ID.make("acme")
const modelID = Model.ID.make("model")
const modelsDev = ModelsDev.Service.of({
get: () =>
Effect.succeed([
{
info: {
id: providerID,
name: "Acme",
activation: "auto",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
},
environment: [],
models: [
{
id: modelID,
modelID,
providerID,
name: "Model",
capabilities: { tools: true, input: [], output: [] },
variants: [],
time: { released: Date.parse("2026-01-01") },
cost: [],
status: "active",
enabled: true,
limit: { context: 128_000, output: 32_000 },
},
],
},
] satisfies readonly ModelsDev.Snapshot[]),
refresh: () => Effect.void,
})
const pluginHost = yield* PluginHost.make(plugins)
yield* ModelsDevPlugin.effect(pluginHost).pipe(Effect.provideService(ModelsDev.Service, modelsDev))
const scope = yield* Scope.make()
yield* catalog
.transform((editor) =>
editor.model.update(providerID, modelID, (model) => {
model.variants ??= []
model.variants.push({ id: Model.VariantID.make("configured") })
}),
)
.pipe(Scope.provide(scope))
expect((yield* catalog.model.get(providerID, modelID))?.variants).toEqual([
{ id: Model.VariantID.make("configured") },
])
yield* Scope.close(scope, Exit.void)
expect((yield* catalog.model.get(providerID, modelID))?.variants).toEqual([])
}),
)
it.effect("keeps the shared models.dev snapshot pristine while catalog transforms mutate records in place", () =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
const providerID = Provider.ID.make("acme")
const modelID = Model.ID.make("gpt-5.4")
const snapshot = [
{
info: {
id: providerID,
name: "Acme",
activation: "auto",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
settings: { baseURL: "https://api.acme.test/v1" },
headers: { "x-acme": "provider" },
},
environment: ["ACME_API_KEY"],
models: [
{
id: modelID,
modelID,
providerID,
name: "GPT-5.4",
settings: { baseURL: "https://models.acme.test/v1" },
capabilities: { tools: true, input: ["text"], output: ["text"] },
variants: [],
time: { released: Date.parse("2026-01-01") },
cost: [],
status: "active",
enabled: true,
limit: { context: 1_050_000, output: 128_000 },
},
],
},
] satisfies readonly ModelsDev.Snapshot[]
const pristine = JSON.stringify(snapshot)
// The plugin receives the same snapshot instance every Location shares.
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
).pipe(
Effect.provideService(
ModelsDev.Service,
ModelsDev.Service.of({ get: () => Effect.succeed(snapshot), refresh: () => Effect.void }),
),
)
// Later plugins mutate nested provider and model records in place, as the Bedrock provider does.
yield* catalog.transform((draft) => {
draft.provider.update(providerID, (provider) => {
if (provider.settings) provider.settings.baseURL = "https://override.acme.test/v1"
if (provider.headers) provider.headers["x-acme"] = "override"
})
draft.model.update(providerID, modelID, (model) => {
if (model.settings) model.settings.baseURL = "https://override.models.acme.test/v1"
model.variants.push({ id: Model.VariantID.make("configured") })
model.capabilities.input.push("image")
})
})
const provider = yield* catalog.provider.get(providerID)
const model = yield* catalog.model.get(providerID, modelID)
expect(provider?.settings?.baseURL).toBe("https://override.acme.test/v1")
expect(provider?.headers).toEqual({ "x-acme": "override" })
expect(model?.settings?.baseURL).toBe("https://override.models.acme.test/v1")
expect(model?.variants).toEqual([{ id: Model.VariantID.make("configured") }])
expect(model?.capabilities.input).toEqual(["text", "image"])
expect(JSON.stringify(snapshot)).toBe(pristine)
}),
)
it.effect("projects normalized models.dev snapshots into the catalog", () =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
const providerID = Provider.ID.make("acme")
const modelID = Model.ID.make("gpt-5.4")
const models = ModelsDev.Service.of({
get: () =>
Effect.succeed([
{
info: {
id: providerID,
name: "Acme",
activation: "auto",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
settings: { baseURL: "https://api.acme.test/v1" },
},
environment: [],
models: [
{
id: modelID,
modelID,
providerID,
name: "GPT-5.4",
family: Model.Family.make("gpt"),
capabilities: { tools: true, input: [], output: [] },
variants: [],
time: { released: Date.parse("2026-01-01") },
cost: [
{
input: Money.USDPerMillionTokens.make(2.5),
output: Money.USDPerMillionTokens.make(15),
cache: {
read: Money.USDPerMillionTokens.zero,
write: Money.USDPerMillionTokens.zero,
},
},
{
tier: { type: "context", size: 272_000 },
input: Money.USDPerMillionTokens.make(3),
output: Money.USDPerMillionTokens.make(18),
cache: {
read: Money.USDPerMillionTokens.make(0.25),
write: Money.USDPerMillionTokens.zero,
},
},
{
tier: { type: "context", size: 200_000 },
input: Money.USDPerMillionTokens.make(5),
output: Money.USDPerMillionTokens.make(22.5),
cache: {
read: Money.USDPerMillionTokens.make(0.5),
write: Money.USDPerMillionTokens.zero,
},
},
],
status: "active",
enabled: true,
limit: { context: 1_050_000, input: 922_000, output: 128_000 },
},
{
id: Model.ID.make("gpt-5.4-fast"),
modelID,
providerID,
name: "GPT-5.4 Fast",
family: Model.Family.make("gpt"),
package: Provider.aisdk("@ai-sdk/openai-compatible"),
settings: { baseURL: "https://api.acme.test/v1" },
headers: { "x-mode": "fast" },
body: { service_tier: "priority" },
capabilities: { tools: true, input: [], output: [] },
variants: [],
time: { released: Date.parse("2026-01-01") },
cost: [
{
input: Money.USDPerMillionTokens.make(5),
output: Money.USDPerMillionTokens.make(30),
cache: {
read: Money.USDPerMillionTokens.make(0.5),
write: Money.USDPerMillionTokens.zero,
},
},
{
tier: { type: "context", size: 272_000 },
input: Money.USDPerMillionTokens.make(3),
output: Money.USDPerMillionTokens.make(18),
cache: {
read: Money.USDPerMillionTokens.make(0.25),
write: Money.USDPerMillionTokens.zero,
},
},
{
tier: { type: "context", size: 200_000 },
input: Money.USDPerMillionTokens.make(5),
output: Money.USDPerMillionTokens.make(22.5),
cache: {
read: Money.USDPerMillionTokens.make(0.5),
write: Money.USDPerMillionTokens.zero,
},
},
],
status: "active",
enabled: true,
limit: { context: 1_050_000, input: 922_000, output: 128_000 },
},
],
},
] satisfies readonly ModelsDev.Snapshot[]),
refresh: () => Effect.void,
})
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
).pipe(Effect.provideService(ModelsDev.Service, models))
const base = yield* catalog.model.get(providerID, Model.ID.make("gpt-5.4"))
const fast = yield* catalog.model.get(providerID, Model.ID.make("gpt-5.4-fast"))
expect(base?.variants).toEqual([])
expect(base?.body).toBeUndefined()
expect(fast).toMatchObject({
id: "gpt-5.4-fast",
modelID: "gpt-5.4",
providerID: "acme",
name: "GPT-5.4 Fast",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
settings: { baseURL: "https://api.acme.test/v1" },
headers: { "x-mode": "fast" },
body: { service_tier: "priority" },
variants: [],
})
expect(fast?.cost).toEqual([
{
input: Money.USDPerMillionTokens.make(5),
output: Money.USDPerMillionTokens.make(30),
cache: {
read: Money.USDPerMillionTokens.make(0.5),
write: Money.USDPerMillionTokens.zero,
},
},
{
tier: { type: "context", size: 272_000 },
input: Money.USDPerMillionTokens.make(3),
output: Money.USDPerMillionTokens.make(18),
cache: {
read: Money.USDPerMillionTokens.make(0.25),
write: Money.USDPerMillionTokens.zero,
},
},
{
tier: { type: "context", size: 200_000 },
input: Money.USDPerMillionTokens.make(5),
output: Money.USDPerMillionTokens.make(22.5),
cache: {
read: Money.USDPerMillionTokens.make(0.5),
write: Money.USDPerMillionTokens.zero,
},
},
])
}),
)
it.effect("omits deprecated models from the catalog", () =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
const providerID = Provider.ID.make("acme")
const activeID = Model.ID.make("current")
const deprecatedID = Model.ID.make("legacy")
const model = {
modelID: activeID,
providerID,
name: "Current",
capabilities: { tools: true, input: [], output: [] },
variants: [],
time: { released: Date.parse("2026-01-01") },
cost: [],
status: "active",
enabled: true,
limit: { context: 128_000, output: 32_000 },
} satisfies Omit<Model.Info, "id">
const snapshots = [
{
info: {
id: providerID,
name: "Acme",
activation: "auto",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
},
environment: [],
models: [
{ id: activeID, ...model },
{
id: deprecatedID,
...model,
modelID: deprecatedID,
name: "Legacy",
status: "deprecated" as const,
},
],
},
] satisfies readonly ModelsDev.Snapshot[]
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
).pipe(
Effect.provideService(
ModelsDev.Service,
ModelsDev.Service.of({
get: () => Effect.succeed(snapshots),
refresh: () => Effect.void,
}),
),
)
expect(yield* catalog.model.get(providerID, activeID)).toBeDefined()
expect(yield* catalog.model.get(providerID, deprecatedID)).toBeUndefined()
}),
)
it.effect("registers key methods for providers with environment variables", () =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
)
expect(yield* integrations.list()).toEqual([
Integration.Info.make({
id: Integration.ID.make("acme"),
name: "Acme",
methods: [
{ type: "key" },
{
type: "env",
names: ["ACME_API_KEY"],
},
],
connections: [],
}),
])
}).pipe(Effect.provide(models(path.join(import.meta.dir, "fixtures", "models-dev.json")))),
)
it.effect("preserves provider and model URL templates in the catalog", () =>
withEnv(
{
ACME_HOST: "api.acme.test",
ACME_MODEL_PATH: undefined,
UNDECLARED_HOST: "private.example",
},
() =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
const providerID = Provider.ID.make("acme")
const modelID = Model.ID.make("gpt-5.4")
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
).pipe(
Effect.provideService(
ModelsDev.Service,
ModelsDev.Service.of({
get: () =>
Effect.succeed([
{
info: {
id: providerID,
name: "Acme",
activation: "auto",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
settings: { baseURL: "https://${ACME_HOST}/${UNDECLARED_HOST}/v1" },
},
environment: ["ACME_HOST", "ACME_MODEL_PATH", "ACME_API_KEY"],
models: [
{
id: modelID,
modelID,
providerID,
name: "GPT-5.4",
settings: { baseURL: "https://${ACME_HOST}/${ACME_MODEL_PATH}/v1" },
capabilities: { tools: true, input: [], output: [] },
variants: [],
time: { released: Date.parse("2026-01-01") },
cost: [],
status: "active",
enabled: true,
limit: { context: 1_050_000, output: 128_000 },
},
],
},
] satisfies readonly ModelsDev.Snapshot[]),
refresh: () => Effect.void,
}),
),
)
expect((yield* catalog.provider.get(providerID))?.settings?.baseURL).toBe(
"https://${ACME_HOST}/${UNDECLARED_HOST}/v1",
)
expect((yield* catalog.model.get(providerID, modelID))?.settings?.baseURL).toBe(
"https://${ACME_HOST}/${ACME_MODEL_PATH}/v1",
)
}),
),
)
it.effect("copies model request bodies without reinterpreting literal __proto__ keys", () =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
const providerID = Provider.ID.make("acme")
const modelID = Model.ID.make("gpt-5.4")
// A JSON body may legitimately contain a "__proto__" key; both copy stages must keep it as an own property.
const body = JSON.parse('{"__proto__":{"service_tier":"priority"},"keep":true}') as Record<string, unknown>
const snapshot = {
info: {
id: providerID,
name: "Acme",
activation: "auto",
package: Provider.aisdk("@ai-sdk/openai-compatible"),
},
environment: [],
models: [
{
id: modelID,
modelID,
providerID,
name: "GPT-5.4",
capabilities: { tools: true, input: [], output: [] },
variants: [],
time: { released: Date.parse("2026-01-01") },
cost: [],
status: "active",
enabled: true,
limit: { context: 1_050_000, output: 128_000 },
body,
},
],
} satisfies ModelsDev.Snapshot
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
).pipe(
Effect.provideService(
ModelsDev.Service,
ModelsDev.Service.of({ get: () => Effect.succeed([snapshot]), refresh: () => Effect.void }),
),
)
const copied = (yield* catalog.model.get(providerID, modelID))?.body
expect(copied).not.toBe(body)
expect(Object.hasOwn(copied ?? {}, "__proto__")).toBe(true)
expect(Object.keys(copied ?? {})).toEqual(["__proto__", "keep"])
expect(JSON.stringify(copied)).toBe(JSON.stringify(body))
expect(copied).not.toHaveProperty("service_tier")
}),
)
it.effect("omits legacy provider aliases", () =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
const snapshots = [
["azure", "Azure", "AZURE_API_KEY", "@ai-sdk/azure"],
["azure-cognitive-services", "Azure Cognitive Services", "AZURE_COGNITIVE_SERVICES_API_KEY", "@ai-sdk/azure"],
["google-vertex", "Google Vertex", "GOOGLE_APPLICATION_CREDENTIALS", "@ai-sdk/google-vertex"],
[
"google-vertex-anthropic",
"Google Vertex Anthropic",
"GOOGLE_APPLICATION_CREDENTIALS",
"@ai-sdk/google-vertex/anthropic",
],
].map(([id, name, environment, packageName]) => ({
info: {
id: Provider.ID.make(id),
name,
activation: "auto",
package: Provider.aisdk(packageName),
},
environment: id === "azure" ? ["AZURE_RESOURCE_NAME", environment] : [environment],
models: [],
})) satisfies readonly ModelsDev.Snapshot[]
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
).pipe(
Effect.provideService(
ModelsDev.Service,
ModelsDev.Service.of({
get: () => Effect.succeed(snapshots),
refresh: () => Effect.void,
}),
),
)
expect(yield* catalog.provider.get(Provider.ID.azure)).toBeDefined()
expect(yield* catalog.provider.get(Provider.ID.make("google-vertex"))).toBeDefined()
expect(yield* catalog.provider.get(Provider.ID.make("azure-cognitive-services"))).toBeUndefined()
expect(yield* catalog.provider.get(Provider.ID.make("google-vertex-anthropic"))).toBeUndefined()
expect(yield* integrations.get(Integration.ID.make("azure"))).toBeDefined()
expect(yield* integrations.get(Integration.ID.make("azure"))).toMatchObject({
methods: [{ type: "key" }, { type: "env", names: ["AZURE_API_KEY", "AZURE_COGNITIVE_SERVICES_API_KEY"] }],
})
expect(yield* integrations.get(Integration.ID.make("google-vertex"))).toBeDefined()
expect(yield* integrations.get(Integration.ID.make("azure-cognitive-services"))).toBeUndefined()
expect(yield* integrations.get(Integration.ID.make("google-vertex-anthropic"))).toBeUndefined()
expect(ProviderPlugins.map((plugin) => plugin.id)).not.toContain("opencode.provider.azure.cognitive.services")
expect(ProviderPlugins.map((plugin) => plugin.id)).not.toContain("opencode.provider.google.vertex.anthropic")
}),
)
it.effect("advertises only key-bearing Google Vertex environment variables", () =>
Effect.gen(function* () {
const integrations = yield* Integration.Service
const catalog = yield* Catalog.Service
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
).pipe(
Effect.provideService(
ModelsDev.Service,
ModelsDev.Service.of({
get: () =>
Effect.succeed([
{
info: {
id: Provider.ID.make("google-vertex"),
name: "Google Vertex",
activation: "auto",
package: Provider.aisdk("@ai-sdk/google-vertex"),
},
environment: ["GOOGLE_VERTEX_PROJECT", "GOOGLE_VERTEX_LOCATION", "GOOGLE_APPLICATION_CREDENTIALS"],
models: [],
},
] satisfies readonly ModelsDev.Snapshot[]),
refresh: () => Effect.void,
}),
),
)
// Vertex authenticates through ADC; project, location, and the credentials
// file path are configuration, not API keys.
expect(yield* integrations.get(Integration.ID.make("google-vertex"))).toMatchObject({
methods: [{ type: "key" }, { type: "env", names: ["GOOGLE_VERTEX_API_KEY"] }],
})
}),
)
it.effect("converts reasoning options into settings variants", () =>
Effect.gen(function* () {
const catalog = yield* Catalog.Service
const integrations = yield* Integration.Service
yield* ModelsDevPlugin.effect(
host({
catalog: catalogHost(catalog),
integration: integrationHost(integrations),
}),
)
const model = yield* catalog.model.get(Provider.ID.openai, Model.ID.make("gpt-reasoning"))
expect(model?.variants?.map((variant) => variant.id)).toEqual([
Model.VariantID.make("low"),
Model.VariantID.make("high"),
])
expect(model?.variants).toContainEqual({
id: Model.VariantID.make("low"),
settings: {
reasoningEffort: "low",
reasoningSummary: "auto",
include: ["reasoning.encrypted_content"],
},
})
expect(model?.variants).toContainEqual({
id: Model.VariantID.make("high"),
settings: {
reasoningEffort: "high",
reasoningSummary: "auto",
include: ["reasoning.encrypted_content"],
},
})
const mode = yield* catalog.model.get(Provider.ID.openai, Model.ID.make("gpt-reasoning-high"))
expect(mode).toMatchObject({
id: "gpt-reasoning-high",
name: "GPT Reasoning High",
headers: { "x-mode": "high" },
body: { service_tier: "priority" },
})
expect(mode?.variants?.map((variant) => variant.id)).toEqual([
Model.VariantID.make("low"),
Model.VariantID.make("high"),
])
const pro = yield* catalog.model.get(Provider.ID.openai, Model.ID.make("gpt-reasoning-pro"))
expect(pro).toMatchObject({
id: "gpt-reasoning-pro",
body: { reasoning: { mode: "pro" } },
})
const budgetModel = yield* catalog.model.get(Provider.ID.anthropic, Model.ID.make("claude-budget"))
expect(budgetModel?.variants).toContainEqual({
id: Model.VariantID.make("high"),
settings: { thinking: { type: "enabled", budgetTokens: 16000 } },
})
expect(budgetModel?.variants).toContainEqual({
id: Model.VariantID.make("max"),
settings: { thinking: { type: "enabled", budgetTokens: 31999 } },
})
const anthropicEffortModel = yield* catalog.model.get(Provider.ID.anthropic, Model.ID.make("claude-opus-4.7"))
expect(anthropicEffortModel?.variants).toEqual([
{ id: Model.VariantID.make("none"), settings: { thinking: { type: "disabled" } } },
{
id: Model.VariantID.make("low"),
settings: { thinking: { type: "adaptive", display: "summarized" }, effort: "low" },
},
])
const anthropicToggleModel = yield* catalog.model.get(Provider.ID.anthropic, Model.ID.make("claude-toggle"))
expect(anthropicToggleModel?.variants).toEqual([
{ id: Model.VariantID.make("none"), settings: { thinking: { type: "disabled" } } },
{
id: Model.VariantID.make("thinking"),
settings: { thinking: { type: "adaptive", display: "summarized" } },
},
])
const opus45 = yield* catalog.model.get(Provider.ID.anthropic, Model.ID.make("claude-opus-4-5"))
expect(opus45?.variants).toEqual([
{ id: Model.VariantID.make("low"), settings: { effort: "low" } },
{ id: Model.VariantID.make("high"), settings: { effort: "high" } },
])
const grok = yield* catalog.model.get(Provider.ID.make("xai"), Model.ID.make("grok-4.5"))
expect(grok?.variants).toEqual(
["low", "medium", "high"].map((id) => ({
id: Model.VariantID.make(id),
settings: { reasoningEffort: id },
})),
)
const minimax = yield* catalog.model.get(Provider.ID.make("opencode-go"), Model.ID.make("minimax-m3"))
expect(minimax?.variants).toEqual([
{ id: Model.VariantID.make("none"), settings: { thinking: { type: "disabled" } } },
{
id: Model.VariantID.make("thinking"),
settings: { thinking: { type: "adaptive", display: "summarized" } },
},
])
const toggle = yield* catalog.model.get(Provider.ID.make("alibaba"), Model.ID.make("toggle-only"))
expect(toggle?.variants).toEqual([
{ id: Model.VariantID.make("none"), settings: { enableThinking: false } },
{ id: Model.VariantID.make("thinking"), settings: { enableThinking: true } },
])
const combined = yield* catalog.model.get(Provider.ID.make("alibaba"), Model.ID.make("toggle-budget"))
expect(combined?.variants).toEqual([
{ id: Model.VariantID.make("none"), settings: { enableThinking: false } },
{
id: Model.VariantID.make("high"),
settings: { enableThinking: true, thinkingBudget: 8000 },
},
{
id: Model.VariantID.make("max"),
settings: { enableThinking: true, thinkingBudget: 16000 },
},
])
const gateway = yield* catalog.model.get(Provider.ID.make("vercel"), Model.ID.make("alibaba/qwen-toggle"))
expect(gateway?.variants).toEqual([
{ id: Model.VariantID.make("none"), settings: { enableThinking: false } },
{
id: Model.VariantID.make("high"),
settings: { enableThinking: true, thinkingBudget: 8000 },
},
{
id: Model.VariantID.make("max"),
settings: { enableThinking: true, thinkingBudget: 16000 },
},
])
const gatewayNova = yield* catalog.model.get(Provider.ID.make("vercel"), Model.ID.make("amazon/nova-2-lite"))
expect(gatewayNova?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: { additionalModelRequestFields: { reasoningConfig: { type: "disabled" } } },
},
{
id: Model.VariantID.make("low"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "low" } },
},
{
id: Model.VariantID.make("high"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "high" } },
},
])
const gatewayFallback = yield* catalog.model.get(
Provider.ID.make("vercel"),
Model.ID.make("deepseek/deepseek-toggle"),
)
expect(gatewayFallback?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: { reasoning: { enabled: false } },
},
{
id: Model.VariantID.make("low"),
settings: { reasoningEffort: "low" },
},
{
id: Model.VariantID.make("high"),
settings: { reasoningEffort: "high" },
},
])
const openrouter = yield* catalog.model.get(Provider.ID.make("openrouter"), Model.ID.make("openrouter-toggle"))
expect(openrouter?.variants).toEqual([
{ id: Model.VariantID.make("none"), settings: { reasoning: { enabled: false } } },
{ id: Model.VariantID.make("thinking"), settings: { reasoning: { enabled: true } } },
])
const google = yield* catalog.model.get(Provider.ID.make("google"), Model.ID.make("gemini-2.5-flash"))
expect(google?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: { thinkingConfig: { includeThoughts: false, thinkingBudget: 0 } },
},
{
id: Model.VariantID.make("high"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 8000 } },
},
{
id: Model.VariantID.make("max"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 16000 } },
},
])
const vertex = yield* catalog.model.get(Provider.ID.make("google-vertex"), Model.ID.make("gemini-2.5-flash-lite"))
expect(vertex?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: { thinkingConfig: { includeThoughts: false, thinkingBudget: 0 } },
},
{
id: Model.VariantID.make("high"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 8000 } },
},
{
id: Model.VariantID.make("max"),
settings: { thinkingConfig: { includeThoughts: true, thinkingBudget: 16000 } },
},
])
const bedrock = yield* catalog.model.get(
Provider.ID.make("amazon-bedrock"),
Model.ID.make("amazon.nova-2-lite-v1:0"),
)
expect(bedrock?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: { additionalModelRequestFields: { reasoningConfig: { type: "disabled" } } },
},
{
id: Model.VariantID.make("low"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "low" } },
},
{
id: Model.VariantID.make("high"),
settings: { reasoningConfig: { type: "enabled", maxReasoningEffort: "high" } },
},
])
const sapGemini = yield* catalog.model.get(Provider.ID.make("sap-ai-core"), Model.ID.make("gemini-2.5-flash"))
expect(sapGemini?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: { modelParams: { thinkingConfig: { includeThoughts: false, thinkingBudget: 0 } } },
},
{
id: Model.VariantID.make("high"),
settings: { modelParams: { thinkingConfig: { includeThoughts: true, thinkingBudget: 8000 } } },
},
{
id: Model.VariantID.make("max"),
settings: { modelParams: { thinkingConfig: { includeThoughts: true, thinkingBudget: 16000 } } },
},
])
const sapNova = yield* catalog.model.get(Provider.ID.make("sap-ai-core"), Model.ID.make("amazon--nova-lite"))
expect(sapNova?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: {
modelParams: { additionalModelRequestFields: { thinking: { type: "disabled" } } },
},
},
{
id: Model.VariantID.make("low"),
settings: {
modelParams: { additionalModelRequestFields: { output_config: { effort: "low" } } },
},
},
{
id: Model.VariantID.make("high"),
settings: {
modelParams: { additionalModelRequestFields: { output_config: { effort: "high" } } },
},
},
])
const sapCohere = yield* catalog.model.get(
Provider.ID.make("sap-ai-core"),
Model.ID.make("cohere--command-a-reasoning"),
)
expect(sapCohere?.variants).toEqual([
{
id: Model.VariantID.make("none"),
settings: { modelParams: { thinking: { type: "disabled" } } },
},
{
id: Model.VariantID.make("low"),
settings: { modelParams: { reasoning_effort: "low" } },
},
{
id: Model.VariantID.make("high"),
settings: { modelParams: { reasoning_effort: "high" } },
},
])
const sapAnthropicEffort = yield* catalog.model.get(
Provider.ID.make("sap-ai-core"),
Model.ID.make("anthropic--claude-4.7-opus"),
)
expect(sapAnthropicEffort?.variants).toEqual([
{
id: Model.VariantID.make("low"),
settings: {
modelParams: {
additionalModelRequestFields: {
thinking: { type: "adaptive", display: "summarized" },
output_config: { effort: "low" },
},
},
},
},
])
const sapAnthropicBudget = yield* catalog.model.get(
Provider.ID.make("sap-ai-core"),
Model.ID.make("anthropic--claude-4-sonnet"),
)
expect(sapAnthropicBudget?.variants).toEqual([
{
id: Model.VariantID.make("high"),
settings: {
modelParams: {
additionalModelRequestFields: { thinking: { type: "enabled", budget_tokens: 8000 } },
},
},
},
{
id: Model.VariantID.make("max"),
settings: {
modelParams: {
additionalModelRequestFields: { thinking: { type: "enabled", budget_tokens: 16000 } },
},
},
},
])
}).pipe(Effect.provide(models(path.join(import.meta.dir, "fixtures", "models-dev-reasoning.json")))),
)
})